Background

Duchenne muscular dystrophy (DMD) is an X-linked recessive hereditary disease and the most frequent lethal inherited disease, affecting 1 in about 3500–5000 newborn males [1]. Mutations in the dystrophin gene (DMD) that disrupt the reading frame and lead to prematurely aborted dystrophin synthesis give rise to DMD. Dystrophin is a critical cytoskeletal structural protein in muscle cells that stabilizes the dystrophin glycoprotein complex (DGC) at the sarcolemma [2]. Loss of functional dystrophin leads to DGC degradation and muscle fiber membrane fragility, resulting in progressive muscle degeneration [3] that leaves patients wheelchair-bound at about the age of 12 [4] and premature death as they reach their twenties. To date, no effective treatments for DMD have been developed.

DMD is the largest known gene with 2.6 million base pairs (bp) comprising 79 exons. Its large size makes it prone to mutations. In-frame DMD mutations generate an internally truncated but partly functional dystrophin protein that results in Becker muscular dystrophy (BMD), which is observed in 1 in about 20,000 male births and has an intermediate to mild clinical manifestation, with some cases being asymptomatic into the seventh decade of life [5]. Efforts have been made to rescue dystrophin expression in DMD by overexpressing a truncated but functional micro-dystrophin-utilizing lentivirus or adeno-associated virus (AAV) [6,7,8] or by inducing the skip** of out-of-frame exons with antisense oligonucleotides [9,10,11] and the read-through of nonsense mutations with small molecules [12]. However, these approaches are constrained by the virus packaging limit and by the failure to permanently restore dystrophin expression due to the lack of genome manipulation.

Clustered regularly interspaced short palindromic repeats (CRISPR) genome editing technology shows promise for the treatment of genetic disorders [Statistical analysis

Data from continuous variables are expressed as mean ± SEM. The means of two groups were compared by Student’s t test, and P < 0.05 was considered statistically significant.

Results

Strategy for CRISPR/Cas9-mediated correction of the DMD gene

The dystrophin protein is mainly composed of four domains, of which the N-terminal, the C-terminal, and the cysteine-rich domains are functionally more important and serve as binding sites of dystrophin and transmembrane protein or sarcolemma. By contrast, much of the central rod domain, which contains the mutational “hot spot,” is not functionally necessary (Fig. 1a). Genome editing to reframe DMD by CRISPR-mediated nonhomologous end joining has been demonstrated as an effective approach for rescuing the expression of functional truncated dystrophin, and it has been shown to result in conversion of DMD to a BMD-like phenotype. Efficiency, applicability, and safety are the major concerns in the clinical application of CRISPR-mediated genome editing. In this study, DMD–MDSCs with deletion of exon 51, which introduces a termination codon in exon 52 and disrupts the open reading frame (ORF) (Fig. 1a), were used as target cells to optimize the strategy for CRISPR/Cas9-mediated gene therapy of DMD. CRISPR gene editing restored the disrupted ORF by inducing large-scale deletion of multiple exons or by introducing insertion or deletion mutations (indels) in exon 50; correlative gRNAs were designed targeting introns 44, 45, 50, 54, and 55 (Additional file 2: Table S1). In the first strategy (∆45–55), exons 45–55 (708 kb; Fig. 1b), which accounted for approximately 63% of the mutations of the DMD gene, were deleted. In the second strategy (∆46–54), exons 46–54 were deleted (334 kb; Fig. 1c) to reframe the mutant DMD, thereby preserving more of the rod domain and making dystrophin with a potentially better function. In the third strategy (INDEL50), the 3′ end of exon 50 was targeted with a single gRNA to restore the downstream reading frame via introduction of indels (Fig. 1d). Compared with the strategies for large fragment deletion, introducing indels improved the efficiency of cutting the genome (only one gRNA).

Fig. 1
figure 1

Structure of human dystrophin protein and strategies for targeting the mutant human DMD gene. a Schematic illustration of the human dystrophin protein structure and the reading frame of the DMD gene derived from a DMD patient harboring a deletion of exon 51 and a termination codon in exon 52 (red asterisk). H, Hinge; R, Repeat; E, Exon. b Strategy I: deletion of exons 45–55 (∆45–55) joins exon 44 to exon 56 and restores the reading frame. E: Exon. c Strategy II: deletion of exons 46–54 (∆46–54) joins exon 45 to exon 55 and restores the reading frame. E, exon. d Strategy III: introduction of random indels in exon 50 (INDEL50) has a chance of restoring the reading frame

Full size image

CRISPR/Cas9-mediated reframing of human mutant DMD gene

gRNAs targeting introns 44 and 55 and exon 50 for the first (∆45–55) and third (INDEL50) strategies were designed on the basis of the screen for DMD targeting as previously reported [20, 21]. To achieve effective deletion of exons 46–54 (∆46–54; Fig. 1c), we generated a dual-fluorescence reporter to rapidly assess the genomic deletion efficiency of Cas9/gRNAs targeting introns 45 and 54. The reporter contained enhanced green fluorescent protein (EGFP) and tdTomato genes separated by DMD introns 45 and 54, which included the gRNA target sequence and flanked by a multi-polyA sequence with a robust transcription termination function (Fig. 2a; Additional file 2: Table S2). CRISPR/Cas9-mediated excision was predicted to remove the multi-polyA cassette to activate tdTomato expression. HEK293 cells were co-electroporated with gRNA targeting DMD intron 45 (gRNA nos. 1–3) and another targeting DMD intron 54 (gRNA nos. 4–6) along with Cas9 and the reporter construct (Additional file 1: Figure S1A). After 48 h, dual fluorescence was detected under a microscope (Fig. 2b), and excision efficiency was evaluated by flow cytometry (Fig. 2c; Additional file 1: Figure S2). Genomic end-point PCR was performed to detect excision of the DMD gene from genomic DNA in the transfected HEK293 cells. The expected genomic deletions were observed for most of the gRNA combinations (Additional file 1: Figure S1B). gRNAs 1–4 showed the highest gene-editing activity according to the results of fluorescence reporter, and these were used for genetic correction of DMD (Additional file 2: Table S3). Sanger sequencing confirmed the joining of introns 45 and 54 mediated by Cas9/gRNA1–4 (Additional file 1: Figure S1C). To test the specificity of Cas9/gRNA-mediated cleavage, we identified 24 potential off-target sites of gRNA1 and gRNA4 via an in silico prediction method and performed deep sequencing of HEK293 cells transfected with Cas9 and individual gRNA expression cassettes. Results showed that off-target mutagenesis occurred at a low frequency (Additional file 1: Figures S1D–S1F; Additional file 2: Table S4).

Fig. 2
figure 2

Screening gRNAs for CRISPR/Cas9-mediated targeting of the DMD gene and reframing of mutant DMD in DMD–MDSCs. a Schematic illustration of gRNAs targeting introns 45 and 54 of the DMD gene (top row). Red and black arrows indicate the target sites of gRNAs and the primer binding sites for the PCR experiments shown in d. Illustration of a dual-fluorescence reporter plasmid (bottom row). b HEK293 cells were co-transfected with the reporter construct and Cas9 (top row) or Cas9/gRNA1–4 (bottom row) by electroporation, and EGFP and tdTomato (Td) expression was observed with a fluorescence microscope 2 days later. Scale bar, 100 μm. c Flow cytometry analysis of HEK293 cells co-transfected with the reporter construct and Cas9/gRNA (n = 2). d Confirmation of the reframing of mutant DMD in DMD–MDSCs by PCR and Sanger sequencing; ∆45–55 yielded an intron 44/intron 55 junction (left column), ∆46–54 yielded an intron 45/intron 54 junction (middle column), and INDEL50 yielded T7E1-cleaved PCR amplicon fragments of the expected sizes (red arrowheads); the red arrow indicates the gRNA target site (right column). M, marker; I, intron; E, exon. e Targeting efficiency of gRNAs in DMD-MDSCs. f The ratio of different types of indels in INDEL50. The numbers of reads of different indel types in three replicates was averaged, and the ratio was calculated. INSERT 3 N+2 and DELETE 3 N+1 were in-frame mutations

Full size image

To reframe the mutant DMD gene in the DMD–MDSCs, the gRNAs for the three targeting strategies (Figs. 1b–d) were separately introduced along with Cas9 into the DMD–MDSCs by electroporation. After 3 days, genomic DNA was extracted from the transfected DMD–MDSCs for analysis. Genomic deletions and indel mutations were assessed by end-point PCR and the Surveyor nuclease assay. The expected deletion products of 708 kb and 334 kb in ∆45–55 and ∆46–54 targeted cells were detected via PCR of the genomic DNA. Sanger sequencing confirmed the reframing of mutant DMD (Fig. 2d; Additional file 1: Figures S3A and S3B). Notably, INDEL50 induced a conversion to three reading frames, with only 1/6 of the mutations restoring the ORF of mutant DMD as previously reported (Additional file 1: Figure S4). The targeting efficiency of gRNAs in the DMD–MDSCs was also detected by high-throughput sequencing of the PCR products of the edited DMD gene (Fig. 2e; Additional file 1: Figure S5; Additional file 2: Table S5). About 6% of the reads included indels from E50-gRNA-edited PCR products, in which 31% of the edited reads were in-framed (Fig. 2f). However, 1–2% of the reframed reads contained stop codons, and about 30% of the reframed reads lost the 5′ splice site (SS) (Additional file 2: Table S6).

Restoration of dystrophin expression by CRISPR/Cas9

To confirm CRISPR/Cas9-mediated restoration of dystrophin expression, we first examined the differentiation potential of the myotubes in the DMD–MDSCs. We differentiated the MDSCs into myotubes in vitro and examined the expression of the myogenic regulatory factors Myogenin, MyoD1, and Desmin, as well as the mature myotube marker myosin heavy chain (MHC), via reverse transcription (RT) PCR and immunofluorescence analysis (Additional file 1: Figure S6). The DMD–MDSCs targeted by the three distinct Cas9/gRNA gene-editing strategies were differentiated into myotubes, and the restoration of dystrophin protein expression was evaluated. Truncated mRNA transcripts were detected in the cells targeted with dual gRNAs (Fig. 3a). Sanger sequencing of the cDNA deletion band revealed the expected fusion of exons 44–56 and 45–55 with the ∆45–55 and ∆46–54 deletion strategies, respectively (Fig. 3b). For the INDEL50 strategy, in which a single gRNA was used to correct the aberrant reading frame by INDEL-mediated frameshifting, T-A cloning and sequencing of cDNA revealed that only 15% (3/20) of the transcripts were in frame (Additional file 1: Figure S7). Restoration of truncated dystrophin protein expression in the cells targeted with the three strategies was confirmed by Western blotting. As shown in Fig. 3c, the size of the truncated protein corresponding to the weight expected in the absence of exons 45–55 (∆45–55) and exons 46–54 (∆46–54) was approximately 361 and 375 kDa, respectively, whereas the molecular weight of full-length dystrophin was 427 kDa in both wild-type (WT) and INDEL50-targeted myotubes, although the latter showed only trace amounts of dystrophin. The intensity of ∆46–54 and ∆45–55 bands was higher than that of INDEL50 band. No dystrophin was detected in the proteins extracted from the DMD myotubes that had not been genetically edited (also see in Additional file 1: Figure S8).

Fig. 3
figure 3

Restoration of dystrophin expression by CRISPR/Cas9. a CRISPR/Cas9-mediated gene editing restored dystrophin mRNA level in myotubes differentiated from DMD–MDSCs. Illustration of primer binding sites (black arrow) for RT-PCR (top row); all detected bands were of the expected size; red arrows indicate bands that were verified by Sanger sequencing in b. M, marker; Un, unedited. b The successful reframing of cDNA in the DMD–MDSCs subjected to three different gene-editing strategies was confirmed by Sanger sequencing. c Western blot analysis of dystrophin expression in targeted (∆45–55, ∆46–54, and INDEL50) or untargeted myotubes differentiated from the DMD–MDSCs; WT MDSCs served as a positive control. Expected molecular weights were 427 kDa for WT and INDEL50, 361 kDa for ∆45–55, and 375 kDa for ∆46–54; MHC served as a loading control. d Representative images of MHC (green) and dystrophin (red, white arrow) expression in targeted (∆45–55, ∆46–54, and INDEL50) and untargeted myotubes differentiated from the DMD–MDSCs as determined by immunocytochemistry; WT MDSCs served as a positive control, and nuclei were stained with DAPI (blue). Scale bar, 100 μm

Full size image

The relative efficiency of the restoration of dystrophin protein expression in the myotubes by CRISPR/Cas9 was evaluated by immunocytochemistry. MHC was expressed in all myotubes. However, although dystrophin protein was detected in the myotubes differentiated from WT–MDSCs, it was absent in those differentiated from the DMD–MDSCs. When CRISPR/Cas9 worked, the expression of dystrophin was restored in the DMD–MDSC myotubes. We calculated the ratio of the number of dystrophin-positive fibers to that of MHC-positive fibers as a measure of in vitro therapeutic efficiency. As shown in Fig. 3d, ∆46–54/Cas9, ∆45–55/Cas9, and INDEL50/Cas9 restored the expression of dystrophin to varying degrees, whereas ∆46–54/Cas9 and ∆45–55/Cas9 were more efficient than INDEL50/Cas9 (also see in Additional file 1: Figure S9).

CRISPR/Cas12a rescues dystrophin expression

Streptococcus pyogenes Cas9 (SpCas9), which is currently the most widely used Cas9 endonuclease, depends on a G-rich protospacer-adjacent motif PAM (NGG) that precludes genome editing of AT-rich regions. Previous studies have demonstrated that Cas12a, an RNA-guided endonuclease that prefers a T-rich PAM, is effective in mammalian genome editing [33, 39]. Thus, we used CRISPR/Cas12a to correct mutant DMD via deletion of exons 46–54 (Fig. 1c). Six Cas12a–gRNAs were designed to target introns 45 and 54 of DMD. The deletion efficiency of Cas12a in HEK293 cells combined with different gRNA pairs was evaluated with a dual-fluorescence reporter and by end-point PCR amplification of endogenous genomic DNA. Cas12a–gRNA pairs 1–4 exhibited the highest gene editing efficiency, and a low off-target frequency was selected for correction of the DMD–MDSCs (Additional file 1: Figures S10 and S11; Additional file 2: Tables S3 and S7).

Expected deletion in genomic DNA and truncated mRNA transcripts were detected in ∆46–54/Cas9- and ∆46–54/Cas12a-targeted MDSCs, and Sanger sequencing confirmed the reframing of mutant DMD (Fig. 4a, b; Additional file 1: Figure S3C). ddPCR was performed to detect the large-scale excision efficiency mediated by Cas9 and Cas12a. Primer sets and probers were designed to target both the excised and the intact DMD gene (Additional file 2: Table S8). ddPCR indicated no significant difference among the three large-scale excision strategies meditated both by Cas9 and Cas12a, and over 2% of the DMD alleles had large-scale deletions (Fig. 4c; Additional file 1: Figure S12). Nanopore sequencing, a long-read sequencing technology, was performed to inspect the spliced DMD fragments and understand the splicing changes mediated by large-scale excision. About 2 kb regions nearby the cutting sites were amplified from the DMD–MDSCs edited by Cas9 or Cas12a with paired gRNAs. Over 50% of the fragments were observed to be exactly spliced on the cutting sites mediated both by Cas9 and Cas12a, whereas Cas12a was more prone to generating spliced fragments with deletions of about 500 bp (Fig. 4d, e; Additional file 1: Figures S13 and S14). The restoration of dystrophin protein expression was detected by Western blotting and immunocytochemistry (Fig. 4f, g; Additional file 1: Figure S15). We also evaluated the efficiency of the three Cas9- and Cas12a-mediated gene-editing strategies via double immunofluorescence labeling with antibodies against MHC and dystrophin and by calculating the ratio of the number of dystrophin-positive myotubes to that of MHC-positive myotubes. ∆46–54/Cas9 and ∆46–54/Cas12a both showed high efficiency (Fig. 4h), whereas DMD gene reframing by INDEL50/Cas9 was inefficient in restoring dystrophin. The efficiency of Cas12a-mediated ∆46–54 was comparable to that of Cas9-mediated ∆46–54.

Fig. 4
figure 4

CRISPR/Cas12a -induced rescue of dystrophin expression. a PCR analysis of dystrophin expression in DMD–MDSCs targeted by Cas9/gRNA or Cas12a/gRNA specific to introns 45 and 54, respectively. All detected bands were of the expected size; the band indicated by a red arrow was verified by Sanger sequencing, which confirmed the junction of segmental introns 45 and 54. M, marker. b RT-PCR analysis of dystrophin expression in myotubes differentiated from the DMD–MDSCs targeted by Cas9/gRNA or Cas12a/gRNA, as shown in a. All detected bands were of the expected size; the band indicated by a red arrow was verified by Sanger sequencing, which confirmed the junction of exons 45 and 55. M, marker. c The editing efficiency of three large-scale excision strategies indicated by ddPCR assays. d Nanopore sequencing reads from edited MDSCs mapped at the spliced DMD genome; cut sites were marked by red lines. E, exon. e The base mutation percentage around the cut sites (40 bp for ∆45–55/Cas9, 65 bp for ∆46–54/Cas9 and ∆46–54/Cas12a). f Western blot analysis of dystrophin expression in targeted (∆46–54/Cas9 and ∆46–54/Cas12a) or untargeted myotubes differentiated from the DMD–MDSCs; WT MDSCs served as a positive control. Expected molecular weights were 427 kDa for WT and 375 kDa for Cas9/∆46–54 and Cas12a/∆46–54. MHC served as a loading control. g Representative images of MHC (green) and dystrophin (red, white arrow) expression in Cas12a/∆46–54-targeted myotubes differentiated from the DMD–MDSCs as determined by immunocytochemistry; nuclei were stained with DAPI (blue). Scale bar, 100 μm. h Representative box plots of the efficiency of the different gene-editing strategies in vitro as indicated by the ratio of the number of dystrophin-positive fibers (dystrophin+) to that of MHC-positive fibers (MHC+); ∆46–54/Cas9, ∆46–54/ Cas12a, and ∆45–55/Cas9 showed higher efficacy than INDEL50/Cas9 (P < 0.05, n = 10)

Full size image

Off-target and transcriptome analyses

Although CRISPR-mediated gene editing is an effective approach for correcting disease-related gene mutations, mitigating the risk of off-target mutagenesis is critical for its application in gene therapy. Given that in silico prediction has poor sensitivity, high-throughput sequencing was conducted to assess the target specificity of both Cas9/gRNA and Cas12a/gRNA. The genomic DNA from the DMD–MDSCs targeted by Cas9/gRNAs and Cas12a/gRNA was analyzed via whole-exome sequencing. In contrast to the unedited DMD–MDSCs, multiple SNVs and indels were detected in both Cas9- and Cas12a-targeted cells with no significant difference (Additional file 1: Figure S16). Furthermore, 16 sites from ∆45–55/Cas9, 14 sites from ∆46–54/Cas12a, 8 sites from ∆46–54/Cas9, and 4 sites from INDEL50/Cas9 were selected for targeted NGS to validate the off-targets detected by whole-exome sequencing (Additional file 2: Table S9). The results showed a very low proportion of mutant reads in the targeted cells, and it was not significantly different from the controls (Additional file 1: Figures S17–S19). Given that the off-targeting effect of CRISPR is highly variable and hardly detected by whole-genome or whole-exome sequencing, GUIDE-seq [38] technology was conducted to capture all DNA double-stranded breaks (DSBs) generated by Cas9 and Cas12a. As we failed to capture any DSBs in the edited MDSCs, GUIDE-seq was performed using Cas9- or Cas12a-edited HEK293 cells as previously described. Dozens to hundreds of off-targets were detected within Cas9–gRNAs (Additional file 1: Figure S20). Although eight sites from IN54–Cas9–gRNA and one site from IN45–Cas9–gRNA were detected by GUIDE-seq as located in exons, high-throughput sequencing showed that those sites were not mutant in the MDSCs (Additional file 1: Figure S21). Taken together, whole-exome sequencing and GUIDE-seq indicated no detectable off-target activity in the MDSCs.

To confirm the functional rescue of the myotubes derived from the edited DMD–MDSCs, we performed whole transcriptome analysis by RNA sequencing (RNA-seq). Hierarchical clustering (Fig. 5a) and principal component analysis (Additional file 1: Figure S22A) showed that the myotubes edited by large-scale-excision were significantly different from the negative controls but similar to the WT controls. Gene Ontology analysis revealed that transcripts that were highly enriched in the edited myotubes were related to muscle organ development, striated muscle tissue development, and muscle tissue development (Fig. 5b; Additional file 1: Figure S22B), suggesting that the rescued myotubes have a function more closely related to muscle function. In addition, differential expression analysis identified 715, 667, 973, and 51 differentially dysregulated genes in ∆45–55/Cas9, ∆46–54/Cas9, ∆46–54/Cas12a, and INDEL50/Cas9, respectively, compared with the negative controls (Additional file 1: Figure S23A). Together, these four strategies affected 28 genes. However, considering the limited therapeutic effect of INDEL50/Cas9, when it was excluded, the other three strategies affected a total of 469 genes (Additional file 1: Figure S23B), including the upregulated genes FOS and the downregulated genes COL6A1, IGF1, and TGFB1 (Additional file 1: Figures S23C and S23D). The expression of these genes in the edited myotubes was confirmed by Western blot (Additional file 1: Figure S23E). All of those genes are involved in muscle cell proliferation, differentiation, and muscle fibrosis and may affect muscle development and function.

Fig. 5
figure 5

Transcriptome analyses. a Heatmap and hierarchical tree comparing the global transcriptome differences in edited (∆45–55/Cas9, ∆46–54/Cas9, ∆46–54/Cas12a and INDEL50/Cas9, n = 2) and negative control (Cas9 only, mock and unedited, n = 2) myotubes differentiated from the DMD–MDSCs with wild-type (WT, n = 4) myotubes differentiated from normal MDSCs. Red and green color intensities represent gene upregulation and downregulation, respectively. Gene ontology (GO) enrichment results of two clustered gene sets are shown on the right side of the heatmap; −log10 p values were used to produce the bar plots. b Gene Ontology analysis of transcripts in edited myotubes

Full size image

Generation of PDX DMD mouse model

Although CRISPR-mediated correction of the DMD gene has been demonstrated to be highly effective in cultured human cells, mouse models, and even a large animal model (i.e., dog), an important step for the clinical translation of this gene-editing strategy is to demonstrate its efficacy and safety in human muscle fibers in vivo. To this end, we transplanted the DMD–MDSCs into the TA muscle of NSI mice to generate a PDX DMD mouse model harboring human muscle fibers. We first examined changes in the transplanted MDSCs on the basis of luciferase activity. The MDSCs were infected with the lentivirus-expressing luciferase to conveniently monitor survival status in vivo. The NSI mice were treated with cardiotoxin to induce a niche for muscle fiber regeneration from human MDSCs (Additional file 1: Figure S24). On the second day, the MDSCs expressing luciferase were injected into TA muscles treated with cardiotoxin. Photon flow was evident from in vivo luminescence imaging at 4 weeks post transplantation, indicating that human MDSCs had settled in the NSI mice (Fig. 6a; Additional file 1: Figure S25; Additional file 2: Table S10). To further validate the presence of human cells, we performed a reported highly sensitive quantitative PCR assay based on the detection of human mitochondrial DNA, which allows for the detection of as few as one human cell in 10,000 mouse cells [37, 40]. The presence of human cells in the TA muscles injected with the DMD–MDSCs was approximately 2.5% (Fig. 6b). Moreover, human spectrin and Lamin A+C were detected in the muscle sections after 4 weeks (Fig. 6c), indicating that human MDSCs were able to differentiate into mature muscle fibers in mice. As shown in Fig. 6c, the human Lamin A+C-positive nuclei marking all human cells in the transplanted muscle always occurred with human spectrin-positive sarcolemma marking the differentiated human muscle fibers, showing high efficiency of myogenic differentiation of the DMD–MDSCs in vivo.

Fig. 6
figure 6

In vivo differentiation of DMD–MDSCs into muscle fibers. a DMD–MDSCs infected with lentivirus-expressing luciferase were transplanted into recipients, and engraftment was monitored by imaging over a period of 4 weeks. Bioluminescence images of representative injected mice were acquired on the indicated days. b qPCR analysis of human mitochondrial DNA revealed the presence of human cells in PDX DMD mouse following injection of DMD–MDSCs over a period of 4 weeks. Mouse DNA samples (M) and human DNA samples (H), as well as a series of mouse–human cell dilution samples, were run as negative and positive controls to estimate the degree of human cell contribution. The red dotted lines indicated that the proportion of human cells in the TA muscle of PDX DMD was between 2 and 2.5%. c Immunofluorescence staining of laminin (green), human spectrin (red signal in sarcolemma), and human Lamin A+C (red signal in nucleus) in the TA muscle of PDX DMD mice; nuclei were counterstained with DAPI (blue). Scale bar, 100 μm

Full size image

In vivo gene editing of human DMD muscle fibers restores dystrophin expression and function

We next examined the capacity of CRISPR/Cas9- and CRISPR/Cas12a-mediated gene editing to correct the mutant DMD gene in vivo. We first evaluated in vivo gene transfer to muscle fibers mediated by AAV9 vector. At 1 week post infection, EGFP expression was detected in the injected muscle tissue, indicating the robust delivery of AAV9 (Additional file 1: Figure S26). To evaluate the possibility of correcting DMD muscle fibers in vivo, we packaged Cas9/gRNA and Cas12a/gRNA targeting various loci in the human DMD gene into AAV9 vectors and delivered to the TA muscle via intramuscular injection 2 weeks after transplantation of the DMD–MDSCs into the NSI mice. After 4 weeks, human dystrophin expression was detected in the muscle tissues (Fig. 7a; Additional file 1: Figure S27). Considering the robust myogenic differentiation of the DMD–MDSCs in vivo, we calculated the ratio of the number of human dystrophin-positive fibers to that of human Lamin A+C-positive nuclei as a measure of in vivo relative therapeutic efficiency. ∆45–55/Cas9, ∆46–54/Cas9, and ∆46–54/Cas12a were highly efficient in rescuing dystrophin expression in vivo (Fig. 7b), consistent with the in vitro results. We also detected the expected bands in ∆45–55- and ∆46–54-targeted mice via PCR of the genomic DNA extracted from the treated muscles. The results confirmed the genome editing of human muscle fibers in the PDX DMD mice (Fig. 7c). In addition, we examined the expression of the DGC component β-dystroglycan to determine the functionality of the rescued myotubes. Dystrophin and β-dystroglycan were correctly localized in all rescued human muscle fibers (Fig. 7d), demonstrating that DGC function was restored and the DMD muscle fiber membrane was stable.

Fig. 7
figure 7

In vivo editing of human DMD gene restores dystrophin expression and localization in muscle fibers. (A) Immunofluorescence detection of human dystrophin (green signal in the sarcolemma, white arrow), human lamin A+C (green signal in the nucleus), and laminin (red) in the TA muscle of CRISPR-targeted PDX DMD mice; nuclei were counterstained with DAPI (blue). Mice without editing (unedited) served as the negative control. Scale bar, 100 μm. b Representative box plots of the therapeutic efficacy of the different gene-editing strategies in vivo as determined by the ratio of the number of dystrophin-positive fibers (dystrophin+) to that of lamin A+C-positive nuclei (LaminA+C+); ∆46–54/Cas9, ∆46–54/Cas12a, and ∆45–55/Cas9 showed higher efficacy than INDEL50/Cas9 (P < 0.05, n = 8). c Confirmation of the presence of human cells and reframing of mutant DMD in PDX DMD mice by PCR. ∆45–55 yielded an intron 44/intron 55 junction, and ∆46–54 yielded an intron 45/intron 54 junction. hmtDNA, human mitochondrial DNA. d β-Dystroglycan restoration in human muscle fibers treated with different gene-editing strategies. Human dystrophin and β-dystroglycan are visible as green and red signals, respectively; sections were stained with DAPI (blue) to identify nuclei and labeled with an antibody against human Lamin A+C (green or red) to identify human nuclei. Mice without editing (unedited) served as a negative control. Scale bar, 50 μm

Full size image

Discussion

In the present study, we used CRISPR-mediated gene-editing technology to repair functional human muscle fibers in a mouse model of DMD by using patient-derived primary MDSCs. A comparison of various gene-editing strategies mediated by particular gRNAs was performed in the DMD–MDSCs and the PDX model, and highly efficient strategies were identified.

CRISPR, which is adapted from bacterial acquired immune surveillance system, is a useful tool for genome editing and the treatment of various inherited diseases. Previous studies have shown that it can be used to modify the DMD gene in the skeletal myoblasts of DMD patients [20, 41], induced pluripotent stem cells (iPSCs) [21, 22], and in the germ line of DMD mice [42]. The CRISPR/Cas9 system can also enable the rescue of dystrophin protein expression in mdx [23,24,25], mdx4cv [43] and genetically humanized mice [28], and in dogs [27] by using AAV vectors or in vivo electroporation. In all of these cases, CRISPR-mediated gene editing permanently restored dystrophin protein expression to alleviate the clinical symptoms of DMD, yielding a phenotype more closely resembling that of BMD.

Previous studies did not address whether mutant dystrophin in DMD patient-derived muscle cells could be edited in vivo via gene editing. We addressed this question by using a PDX model, which more authentically mimics human pathophysiological processes, response to therapy, and provides a sensitive indicator of in vivo efficacy. PDX mouse models have been established by transplanting seed cells, such as lymphocytes [44], hematopoietic stem cells [44], hepatic cells [45], and cancer cells [46], into immunodeficient mouse. We used human primary MDSCs, which are characterized by high proliferative and myogenic differentiation capacities and low immunogenicity [34], from a DMD patient as the seed cells for heteroplastic transplantation into immunodeficient NSI mice that lack B/T lymphocytes and natural killer cells [47, 48]. The transplanted DMD–MDSCs differentiated into mature muscle fibers in vivo, indicating successful establishment of the PDX DMD mouse model. CRISPR/Cas9 was previously used to target genomic mutations in mouse PDX tumor models [49,50,51] in which cancer cells exhibited unlimited proliferation, which is very different from terminally differentiated muscle cells. The present study is the first to apply CRISPR-mediated genome editing of a specific genomic locus (the DMD gene) in adult human nontumor cells (i.e., MDSCs) for the restoration of protein expression and function in vivo.

Cas9 is currently the most widely used nuclease for gene editing. However, its application is limited by the fact that Cas9-mediated cleavage requires a G-rich PAM at the 3′ end of the protospacer for its editing function, whereas introns harbor T-rich regions. In contrast to Cas9, Cas12a recognizes T-rich PAM sequences at the 5′ end of target DNA sequences, which allows targeting of regulatory regions and introns in AT-rich genomes. This useful property expands the range of disease-related mutations that can be edited with the CRISPR system [52]. Moreover, Cas12a is a highly specific programmable nuclease that can reduce off-target effects compared with Cas9 [39]. Our results revealed that ∆46–54/Cas12a-gRNAs have fewer off-target sites than ∆46–54/Cas9–gRNAs, and the deletion efficiency of Cas12a is comparable to that of Cas9. Thus, Cas12a provides an alternative nuclease for future gene therapy in DMD patients.

The efficiency of gene-editing strategies employed in previous studies is unclear due to the use of immortalized myoblasts or iPSCs derived from DMD patients. Herein, we designed a novel method for reframing mutant DMD by deletion of exons 46–54, which encompasses nearly 60% of DMD mutations, and compared the efficiency of this and previously reported approaches both in vitro and in vivo. ddPCR and high-throughput sequencing of the PCR products of the edited DMD gene indicated that the four strategies we compared follow a same trend of reframing efficiency. However, RNA-seq and immunofluorescence showed significantly low efficiency of INDEL-50 mediated restoration of dystrophin. High-throughput sequencing results from the E50-gRNA target indicated that about 30% of the reframed reads were unable to transcribe the correct dystrophin, which reduced the dystrophin restoration efficiency of the INDEL50 strategy. Thus, highly efficient gRNA for introducing indel-mediated DMD must be designed. gRNAs targeting the intron–exon junction area can substantially affect 5′ or 3′ splice sites and reduce the recovery efficiency of INDEL strategy. These results indicated that the large-fragment deletion strategies we used were efficient in vivo and might meet the requirements of clinical efficacy. However, considering that the targeting efficiency of gRNAs is crucial to restoring dystrophin expression, the comparisons are only valid for these particular gRNAs and cannot be extended to these approaches in general.

Conclusions

In summary, this study provides evidence for the efficacy of in vivo genome editing to correct disruptive mutations and restore dystrophin expression and function in DMD patient-derived muscle fibers. The PDX DMD mouse model can be used to screen potentially effective strategies for clinical application. The continued studies of gene-editing strategies, gene delivery approach, toxicology, and immunology in large animals will provide further insights into the potential application of gene therapy for DMD.